Dual-rate Data-driven Design in the Time Domain

In the model matching design, a controller is determined by minimizing a matching error between a reference model with an actual control system. The data-driven method, virtual reference feedback tuning (VRFT), uses a pre-filter to compensate for a matching error between a model-based function and a data-driven function. Designing the pre-filter in the time domain instead of the frequency domain allow a controller to be designed without restricting the type of initial process input. The conventional time-domain based VRFT is designed as a single-rate method, in which all periods are unified. When certain periods are limited due to hardware performance or other factors, performance can be improved by setting unrestricted periods independently of the restricted periods. In this study, the time-domain based VRFT is extended to a dual-rate system, in which the sampling period of a process output differs from the holding period of a process input. Even when the sampling period is limited by sensor performance, computation load, and others, the control performance of the proposed dual-rate method is superior to that of the conventional single-rate method because the process input can be updated more frequently compared with the single-rate method.